TSPACE RESEARCH REPOSITORY tspace.library.utoronto.ca

2015 The Oxford Handbook of Philosophy of Perception Introduction Post-print/Accepted manuscript Mohan Matthen

Mohan, M. (2015). Introduction. In M. Mohan (Ed.), The Oxford Handbook of Philosophy of Perception (1-25). Oxford: Oxford University Press.

Reproduced by permission of Oxford University Press.

HOW TO CITE TSPACE ITEMS Always cite the published version, so the author(s) will receive recognition through services that track citation counts, e.g. Scopus. If you need to cite the page number of the TSpace version (original manuscript or accepted manuscript) because you cannot access the published version, then cite the TSpace version in addition to the published version using the permanent URI (handle) found on the record page.

Oxford Handbook of the Philosophy of Perception

Introduction*

Mohan Matthen

University of Toronto

Abstract Perception is the ultimate source of our knowledge about contingent facts. It

is an extremely important philosophical development that starting in the last quarter

of the twentieth century, philosophers have begun to change how they think of

perception. The traditional view of perception focussed on sensory receptors; it has

become clear, however, that perceptual systems radically transform the output of

these receptors, yielding content concerning objects and events in the external world.

Adequate understanding of this process requires that we think of perception in new

ways—how it operates, the differences among the modalities, and integration of

content provided by the individual senses. Philosophers have developed new analytic

tools, and opened themselves up to new ways of thinking about the relationship of

perception to knowledge. The Oxford Handbook of the Philosophy of Perception is a

collection of entries by leading researchers that reviews these new directions in

philosophical thought. The Introduction to the Handbook reviews the history of the

subject from its beginnings in ancient Greece to the nineteenth century, and the way

that science and philosophy have together produced new conceptions during the last

hundred years. It shows how the new thinking about perception has led to a complex

web of theories.

* I am grateful to Ophelia Deroy and Barry Smith for critically reading through the whole

Introduction. Lynne Godfroy was (as always) very helpful in matters of exposition. Other more specific debts are recorded in other footnotes below.

INTRODUCTION

2

Keywords perception, Plato, Aristotle, scepticism, Stoics, Epicurus, agnosia, evolution

of perception, perceptual specialization, sense data, realism.

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

3

Perception is the ultimate source of knowledge about contingent facts. We know

about our surroundings because we are able to experience them through perception;

we know about scientific phenomena because they are observed. Epistemology is

therefore very much concerned with the evidential value of perception. Analytic

epistemology is concerned with the rational grounding that perception lends belief;

empiricist philosophy of science erects the entire edifice of scientific knowledge on

the back of perceptual observation. The rationality of perceptual grounding is

contested, of course. On one side, it is contested by those, like Hume, who think we

never have reason to believe in contingencies. According to him, we arrive at

contingent beliefs about matters beyond mere sense-impressions by the association

of ideas, and not by reason. On the other side, it is impugned by rationalists like Plato

and Descartes, who think that perception is incoherent—both complain, for example,

that it fails to intimate shape in a way that suffices to ground geometry, the

authoritative science of shape—and far too evanescent to offer genuine and secure

knowledge. Still: even the opposition focuses on a critique or reinterpretation of

observation and its epistemic value.

Given the centrality of perception to epistemology, one would expect that the

philosophical study of perception would be a focal philosophical topic. It has not been:

neither traditional epistemology nor traditional philosophy of science has been

particularly concerned to engage in a determined and scientifically informed

investigation of the nature of perception itself. Both articulate puzzles and theories

that come out of deep and original thinking about the problem of knowledge yoked

to relatively superficial and dogmatic thinking about perception. In the last part of the

twentieth century, this situation began to change. Philosophers began to use sensory

psychology as a source of new insights about the nature of perception. Thanks to

growing interest in perception—how it operates, what it reveals—and the

development of new analytic tools, the philosophy of perception is, once again, a vital

and vibrant area of philosophical inquiry.

INTRODUCTION

4

Taken together, the chapters of this Handbook are an introduction to new

philosophical thinking about perception. This Introduction presents an overview of

some global issues, with the aim of contextualizing perception within broader

philosophical concerns. It does not attempt to summarize or discuss individual

entries. Without exception, these are intellectually sophisticated introductions to

sub-areas and as such they stand alone, requiring no additional exposition here.

Accordingly, individual entries are discussed only when they are directly relevant to

the more synoptic topics taken up in the Introduction, though each is at least

mentioned to show their place in the whole. Thus, the Introduction does not attempt

to touch, even very lightly, on all of the many original and often surprising insights

that readers will find in each and every individual entry. It also suppresses

bibliographical references; these are found in the relevant entries.

I.

Until very recently, and to an extent even now, the epistemologist’s paradigm of

perception remains much unchanged since the seventeenth century. According to this

view, what we directly perceive—the message given to us by perception

unsupplemented by inference from other sources—is a conscious presentation that

closely parallels the excitation of sensory receptors. Call this the Receptoral Image

Model (RIM). RIM takes slightly different forms in psychology and philosophy, as I

shall now explain.

RIM traces back at least as far as Johannes Kepler’s theory of the eye. As David

Hilbert (III.1) explains, the great astronomer came to realise that the lens of the eye

refractively focuses all the light rays arriving from any given external location onto a

single retinal place; thus, it creates an image on the retina. According to Kepler, this

image is what we directly see. This discovery of the retinal image was greatly

impressive to those who followed, though it took more than two hundred years for

the realization to dawn that optics is not enough.

It was not until the early nineteenth century that the eminent German

physiologist, Johannes Müller, came to realize that the optical image is not directly

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

5

the starting point of vision. (It is, rather, the last item in the external causal chain that

links object to perception.) For the optical image that is focussed on the retina to

affect conscious sensation, it must first be converted into a pattern of nerve energy.

The retina is packed with neurons that are activated proportionately to the amount

of light that falls on each; the activity of these neurons determines visual

consciousness, Müller proposed. This is an important advance on Kepler, though it

made little, if any, impact on philosophical theory. Philosophers still show little

awareness that the conversion into nerve energy destroys much of the wavelength-

specific information that is available in the optical image—information that could be

extracted by a spectrometer. (See section VI of the Introduction, below.)

Müller’s realization is the basis for generalizing Kepler’s theory beyond vision.

Corresponding patterns of receptor activation could be assumed, and do indeed exist,

for the other senses—the senses all have specialized receptors that are differentially

activated by environmental energy incident on them. RIM assumes that receptor

activations correspond closely to what we perceive in each modality—the activation

of auditory neurons corresponds to what we hear; the activation of haptic neurons

corresponds to what we feel, and so on. The receptoral images in these other senses

do not have exactly the same properties as the retinal image—the auditory image in

particular is poor in spatial information. Nonetheless, the sensory neurons provide

us with what we now call sensory information. RIM identifies this information with

what philosophers call the perceptual given, in other words, with what we perceive

directly and non-inferentially.

Psychology and philosophy worked with somewhat different versions of RIM.

Psychologists assumed that conscious awareness in each sense modality

corresponds to the receptoral image, and they tended to assume, though less

explicitly, that perception beyond the receptoral image is indirect, or inferred by

learned association. For example, since the visual receptoral image is two-

dimensional, they assumed that direct visual awareness must also be of a two

dimensional array. Perceptual awareness of depth and of three-dimensional

INTRODUCTION

6

objects is inferred. Nineteenth century psychology realized that we possess two

such retinal images, and they tried to work out how the discrepancies between

these images could provide a basis for the inference of depth.

Philosophers made a convergent assumption. In early modern philosophy, and for

a long time after, it was commonly assumed that the perceptual given—what we

directly see, or hear, etc.—is that which is certain or indubitable given our state of

sensory awareness. In the case of vision, two coloured regions might be seen side

by side, but it is uncertain, given visual evidence, how far away each is.

Consequently, philosophers too assumed that depth was not directly given in

vision. Berkeley made this inference explicitly, but it is implicit in Descartes and

others.

The important difference between philosophers and psychologists is that the former

are concerned with rational justification and the latter with physiology and conscious

awareness. However, they arrive at comparable conclusions. Both conclude that

visual awareness of three-dimensionality is indirect, the result of learned association

or (according to some philosophers) rational inference. Psychologists and

philosophers use Kepler’s theory differently, because philosophy is supposed to be a

priori, which means that it cannot use a scientific theory as a foundation for

knowledge. For this reason, philosophers cannot explicitly appeal to Kepler’s optics.

Nevertheless, philosophers as diverse as Descartes and Berkeley commonly used the

retinal image as a heuristic: it serves for them (though not explicitly) as a model of

what we see and as the basis for generalizing to the other senses.

RIM encourages many misconceptions regarding perception; collectively

these misconceptions constitute a traditional view of perception that is slowly

eroding away. RIM implies, for example, that:

Perceptual experience is necessarily unimodal (because the receptors and nerve

energies are). (See Tim Bayne and Charles Spence, V.3, for a re-evaluation.)

Objects can be made to look a different colour simply by shining coloured light on

them (because this changes the colour of light that reaches the visual receptors).

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

7

(Jonathan Cohen discusses the limitations of this notion in V.4, as well as

distortions of shape and size in the optical image; Kathleen Akins and Martin Hahn

discuss the case of colour in IV.3.) Similar assumptions can be made in other

modalities, though it was unusual for them to be explicitly worked out. For

example, it could be assumed that since the activation of auditory receptors is

changed by new sounds, direct auditory awareness of continuing sounds would

be modified by new sounds. Again, it could be assumed (with somewhat greater

empirical justification) that gustatory awareness of bacon would be modified by

taking vinegar into one’s mouth.

Like the retinal array, the visual image is a two-dimensional “colour mosaic”—i.e.,

it consists of a two-dimensional matrix of minimally sized coloured dots—that

does not contain depth information. (How do “coloured dots” match up with the

representation of colour in the system? See Akins and Hahn, IV.3, for discussion.)

Analogously, audition offers us a confused melange of sound that does not directly

inform us of spatially separated sources of sound. (Roberto Casati, IV.1, Jérôme

Dokic, IV.4, and Matthew Nudds, III.2, have relevant discussions.)

Flavour is sensed through the tongue alone, for the only receptors that are

specialized for flavour perception reside in the tongue. (Barry Smith, III.4,

reconceptualises flavour as a multisensory quality, and recounts, in particular,

how wrong it is to think that it is restricted to taste receptors on the tongue—

olfactory receptors are involved too, but in an unusual way.)

Because sensory receptors can in principle be excited without any external

stimulus, perception must be subjectively indistinguishable from hallucination.

Since hallucination is not about external events, perception cannot be so either.

Thus, RIM encourages the idea that ordinary perceptual experience is of sensory

events internal to consciousness, not of the world beyond. (See Baron Reed, I.3,

and Paul Snowdon, I.6, for critical discussion.)

Additionally, philosophers have often assumed that since the receptoral image

is constantly in flux, perception is momentary; temporally extended experience is a

INTRODUCTION

8

fusion of successive receptoral images, and involves memory, which is epistemically

on a different footing. Consequently, they assume that:

Experience of change and movement are due to fused temporal sequences of

momentary experiences of positions or qualities. (See Robin Le Poidevin, IV.5,

about temporally extended perception.)

By the same token, insofar as speech and music are perceived, it is by fusing

experience of a stream of momentary tones. (See Casey O’Callaghan, IV.6, on

speech perception, and Charles Nussbaum, IV.7, on music perception.)

These assumptions are, for the most part, gross oversimplifications; some (such as

those concerning the colour mosaic, flavour, and speech) are completely false.

II.

What explains the dominance and long persistence of momentary RIM in

philosophical theorizing? In large part, the answer is historical. RIM, in particular the

claims (a) that certain aspects of perception, such as depth, are fallibly inferred, and

(b) that hallucination is subjectively indistinguishable from ordinary perceptual

experience (Snowdon, I.6), leads quite naturally to the problem of scepticism (Reed,

I.3). Scepticism is one of the central problems of epistemology, with proponents vying

with opponents who quest for theories of justification and of knowledge that can

withstand the sceptical threat. Philosophers concentrated over the years on the ins

and outs of the sceptical threat to knowledge, leaving unexamined the route by which

they arrived at this dangerous place.

Scepticism is a real philosophical problem, but it does not necessarily rest on

RIM. In fact, as explained earlier, the explicit basis for this and other epistemological

theories is a consideration of the role of inference in situations of uncertainty;

preconceptions regarding sensory receptors play a merely heuristic role.

Nevertheless, theoretical progress in epistemological conceptions of perception was

retarded by RIM, because this framework provided a familiar context for motivating

scepticism in relation to perception. The philosophy of perception has long been

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

9

dominated by the so-called “problem of perception,” the problem of how perception,

which is often misleading about the external world, can nonetheless be a foundation

for knowledge about the external world. Much less effort has gone into figuring out

the nature of perception. To wit: is it really true that direct awareness is as RIM would

have it?

Traditionally, epistemologists took their main problem with regard to

perception to be the uncertainly of beliefs that are based on perception. One might

think, however, that epistemologists should be at least as vitally concerned to

understand how we arrive at ordinary perceptual beliefs—how we get to a belief is,

after all, at least partially independent of why it might be mistaken. Take this very

simple question. Do we recognize a musical beat by internally timing successive

pulses, or do we feel the beat more holistically? This is a clear question about how we

arrive at a belief; it is relevant to whether the perception of musical rhythm depends

entirely on a sense of temporal duration, and whether, if it does, this would show that

it rests on memory. This question is independent of the sceptical question of whether

what we hear is real or merely an illusion, and of the question whether we can ever

be absolutely certain that a piece of music has a particular time signature. It is a

question about the perceptual basis for the belief that a piece of music is a waltz. Is

this belief directly delivered to us by perception, or does it rely on a more complex

calculation?

Considered in this context, the problem regarding the traditional RIM

paradigm of perception is not that it encourages scepticism—there is nothing wrong

with this—but rather that it offers an incomplete and often misleading account of

quotidian perceptual belief and knowledge. It is true that colour mosaics sometimes

simulate ordinary visual perception. This is precisely how object perception and

motion perception is simulated on TV. Nevertheless, RIM is uninformative about the

normal process of forming rational beliefs about objects and their movements in

three-dimensional space. And here it is relevant that the psychological heuristic of

momentary receptoral activation is based on false assumptions. For instance, it turns

out to be false that the visual experience of movement is created by a post-perceptual

INTRODUCTION

10

summation of momentary experiences of objects in successive positions. The fact is

rather that a specialized part of the visual brain detects distal motion (differentiating

it from shifts of the retinal displays that are due to the subject’s own motion) without

the intervention of the subject’s rational acuity in inference. It is also not true that we

perceive objects by summing up retinal colour pixels; the brain has specialized

pattern-detecting mechanisms for this (Roberto Casati, IV.1). As it turns out, our

perception of movement and of objects does not depend on perceiving all of the

temporal or spatial parts of these entities.

The philosophical theory about the uncertainty of inference from perception

to belief could have been, should have been, and was maintained even after the

psychological theory of sensory receptoral images had been long abandoned. This

divergence, however, makes it all the more urgent to give a theory of the formation of

ordinary perceptual beliefs. The best psychological theories of sensory awareness

urge that consciousness presents us with something more substantial than receptoral

arrays. At the same time, it is acknowledged that this sensory given is uncertain. (In

fact, one important way of probing the perceptual given is to study the illusions that

occur in normal perceptual situations.) This gives epistemologists strong motivation

to offer better theories of how we ordinarily justify perceptual beliefs. (Susanna

Siegel and Nicholas Silins, VII.1, discuss reason giving for fallible perceptual belief;

Michael Rescorla, VI. 2, E. Samuel Winer and Michael Snodgrass, VI.3, and John

Kulvicki, VI.4, explain frameworks relevant for posing the problem of the perceptual

given.)

In a similar vein, the model of speech perception implicit in the music-analogy

mentioned earlier gives us a false idea of how we come to know what people around

us are saying (O’Callaghan, chapter IV.6). Phoneme perception is not a summation of

temporally punctate auditory experience; phonemes are temporally extended

(though very brief) sound patterns—phonemes are minimal meaning-bearing units

of spoken language; no part of a phoneme is heard as a speech sound, yet they are so

heard as a whole. By itself, this is proof that perceptual experience is not merely a

summation of temporally punctate sensations.

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

11

As far as music is concerned, rhythm and phrasing is temporally extended (and

of relatively long duration, compared to phonemes) but these too are perceived as

wholes, not as units that the perceiver has to put together by her own agency

(Nussbaum, IV.7). As well, vision is involved in speech perception; perception of

speakers’ facial gestures modifies what we seem to hear. Speech and musical

perception thus contradict the notion that intermodal and cross-temporal integration

are always extra-perceptual mental operations. Auditory perception, in general, is

experience of temporally extended objects—sounds—that often appear to change

(Matthew Nudds, III.2), or at least to have a non-uniform temporal profile. For

instance, we might experience a single voiced melody as ululating; or a siren as rising

in pitch. (These questions about speech and music perception are illuminated by the

considerations about time discussed in Le Poidevin, IV.5.)

Something like this is true of flavour too: when we savour food or wine, there

is an early attack followed by extended finish. Think of spearmint gum: it might start

sweet, then become cool, and finish with a slightly bitter “aftertaste.” This has to do

in part with the dissipation of sugar, and it could be argued that spearmint offers us a

progression of flavour experiences, rather than perception of one temporally

extended flavour. However that might be, such progressions are predictable; they are,

moreover, repeatable, and therefore important in the identification and evaluation of

food and wine. In short, they are ecologically salient. (Barry Smith, III.4, discusses.)

Expert tasters become able to use such temporal profiles to refine expert

discriminations by perceptual learning. (Rob Goldstone and Lisa Byrge, VII.2, provide

a general introduction to perceptual learning.)

The scientific study of perception was, until early in the twentieth century,

rooted in many of the same philosophical assumptions that led to the wrong

assumptions recounted in the preceding section (and many others). For example,

visual perception was thought, in the nineteenth century, to consist first in the

transference of retinal images to the primary visual cortex, and then the extraction of

information from this so-called “cortical retina” by exploiting associations between

retinal stimulations and distal features. This is a model that applies to the brain a not

INTRODUCTION

12

very much modified version of Berkeley’s theory of vision. This model has two fatal

weaknesses. First, it misconstrues the nature of the retinal image: the relevant entity

is not the image thrown by the lens, but the firing of sensory neurons caused by this

pattern of light. Secondly, the model does not transfer smoothly to other modalities.

For example, the auditory cortex is a frequency-intensity mapping, not a spatial image.

That is, the activation-level of different regions of this cortex corresponds to energy

levels associated with a particular frequency, not with energy levels coming from a

particular spatial direction.

III.

The laboratory study of perception began in the middle of the nineteenth century.

(See Gary Hatfield, I.5.) At first, it was dominated by the RIM paradigm: Müller held

that the retinal image was transmitted to consciousness by the physical action of

nerves. One of the great controversies of the late nineteenth century was the battle

between Ewald Hering and Hermann von Helmholtz about the extraction of depth

information from the disparity of the two retinal images. Helmholtz thought that the

perception of three-dimensional space was learned by the association of ideas;

Hering was more of a nativist.

Science does not stand still. Gradually new discoveries began to widen the

research focus.

In the Russo-Japanese war and the First World War, surgery had progressed to

the point where soldiers who had been shot in the head could survive; often they

survived with severe but localized brain lesions along the path of the bullet that

penetrated their skulls. Starting in the early part of the twentieth century, many

specialized perceptual deficits were discovered in patients with such lesions,

some produced by injury, others by other adverse events, such as strokes or

prolonged hypoxia. It was found, for example, that some people with normal

visual acuity with respect to colours and lines were nonetheless unable to

recognize forms that are composed of the lines they could see normally: familiar

objects, such as shapes, faces, places, motion, speech, and alphanumeric symbols.

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

13

These deficits are specialized—for example, the inability to recognize faces does

not predict the inability to recognize shapes, and vice versa.

The perceptual deficits just mentioned are known as “agnosias”—they are

recognitional failures with regard to a restricted domain of “higher-level” (or

composed) objects and features sitting on top of normal acuity with regard to

their “lower-level” components. The agnosias directly contradict the Receptoral

Image Theory. They indicate perceptual deficits in the absence of receptoral

defects. The existence of agnosia shows that awareness of objects is not entirely

dependent on awareness of the parts out of which these objects are constructed.

Agnosia cannot be a failure of inference as such, at least not if inference is

construed as a general purpose capacity, since each agnosia is domain specific—

a patient who fails to recognize faces may have no difficulty discerning motion and

vice versa. Each agnosia is associated with a brain lesion in a specific area; each

highlights a part of the brain that is specialized to extract from receptoral images

content about a particular kind of higher-level perceptual object.

The agnosias indicate what is known as modular function in perception—

separated processes for the extraction of distinct perceptual features (Ophelia

Deroy, VI.5). As time passed, it became clear that there are separate processes for

the extraction of even the lowest level features, for example colour and form in

vision.

Psychological evidence for awareness of higher-level perceptual objects was

accumulated by the Gestalt psychologists, who articulated principles of object

awareness (Hatfield, I.5 and Casati, IV.1). They demonstrated that certain types of

displays result in seeming awareness of objects, while others, though very similar,

either do not or result in awareness of very differently shaped or configured

objects. It can be inferred that object-awareness is not learned by relations of

association among receptoral arrays.

Along similar lines, Albert Michotte demonstrated that certain very simple types

of temporal sequence have the look of causal connectedness—for example, one in

which a simple shape (such as a square) approaches another simple shape, stops

INTRODUCTION

14

when the two touch, at which point the second shape starts moving. Perceivers

typically find it hard to resist the appearance of causation (and even of animacy)

in such displays. This shows that Hume was wrong to say we have no impression

of “power,” and that the appearance of causality is nothing but a projection onto

displays of the mind’s propensity to infer one event from another. There is a

primitive impression of causal connection.

Single neuronal-cell electrical recordings and fMRI imaging demonstrated that

certain brain areas are active when certain types of higher-level object are

presented to perceivers: for example, colour (the fusiform gyrus), motion (visual

area 5), faces (the fusiform face area), places (the parahippocampal place area),

speech (Wernicke’s area), etc. This bolsters the conclusions drawn from agnosias

above, inasmuch as it shows that specialized neuroanatomical structures are

dedicated to extracting content about higher-level objects of specific kinds. (Deroy,

VI.1 discusses the significance of the anatomical localization of this kind of

function; see also Hilbert, III.1.)

Starting in the last two decades of the twentieth century, greater attention has

been paid to multisensory integration (Bayne and Spence, V.1; Smith, III.4,

O’Callaghan, IV.6). For example, when two simple shapes move along straight

intersecting lines, they are seen as passing each other, describing an X. However,

when a sound such as a pop or beep is played at the moment of intersection, the

two shapes are seen as bouncing off one another (rather than as continuing on

along their own prior trajectories undisturbed). (This is as a multisensory version

of Michotte’s experiments on the perception of causation.) Again, when a subject’s

hand is hidden from view and stroked with a brush, while a clearly visible rubber

toy hand is synchronously stroked with a brush, the feeling of stroking is spatially

shifted to the visible rubber hand. This is, again, an integration of sensory

stimulations in two modalities resulting in a single unified percept. Additionally,

subjects report feeling that the rubber hand is their own, so there is also an

involvement of interoception. (See Frédérique de Vignemont and Olivier Massin,

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

15

III.3; Alisa Mandrigin and Evan Thompson, IV.8; Brendan Ritchie and Peter

Carruthers, III.5 for discussion of the rubber hand illusion.)

These findings indicate that perception is not a simply a matter of receiving

external influences, but is rather a process which filters and analyses incoming data

in a search for indications of significant external occurrences and states of affairs.

Moreover, they indicate that consciousness can be perceptual; it is awareness of

external objects and processes, not merely a reproduction of the state of our

receptors within our bodies. Thus, RIM is increasingly giving way to the idea that

perception presents us with a rich scene: objects of many sorts with properties that

do not directly affect the sensory receptors. We literally and directly see objects, faces,

places, and shapes; we hear melodies, voices, and phonemes, once again directly and

not by painstakingly piecing them together by the use of learned experiential

associations. We sense the location of events by both touch and sight working

together; the modalities are not correlated just by learned associations.

Neuroscience and psychology are not the only sources of models for

perception. Functional models treat perceptual systems as performing a “task”

without being specific about the physical means by which the task is performed. One

particularly fruitful idea in this arena has been to treat of sensory receptor response

as data, and the task as data processing with the aim of providing the organism with

the means to respond productively to a changing environment. Note that data here

are abstractly conceived. New analytic tools have also come to the forefront in the last

few years to substantiate this conception. The models that are employed to

understand perception have a significant constraint. Sensory data processing is

“analogue,” in the sense that the system (a) has states that causally respond to

sensory inputs, and (b) vary in a roughly continuous way with firing rates of neurons

etc. Consequently, digital models that are widely used to model thought processes are

of limited utility in the perceptual domain.

The following have become entrenched in philosophical thinking:

INTRODUCTION

16

Bayesian reasoning from probabilities, which neural sensory processing is found

widely to implement. (Michael Rescorla writes about this in VI.2)

Signal detection theory, which provides a framework for understanding how

conscious perception is influenced by the context of inquiry, and perhaps by

voluntary attention. (E. Samuel Winer and Michael Snodgrass, VI.3)

Information theory, which gives an account of what kind of use can be made of

environmental signals to adjust to the events from which they emanate (John

Kulvicki, VI.4)

Other analytic tools either are emerging or have receded from prominence—

predictive coding is a candidate for future prominence, while connectionism seems

to have faded; the current status of dynamic systems theory is clouded. The

philosophy of perception tends to be conservatively selective in its attention to such

structures, being more in tune with cognitive neuroscience, broadly speaking, than

with the abstract mathematics of data processing. No doubt, this has a lot to do with

the history of the subject; there are established problem areas in the field that arise

from thinking about psychological function and neurological implementation.

Abstract modelling has been slower to yield fruitful approaches to established

philosophical problem areas—this could be the fault of philosophers, of course—the

above-mentioned tools being exceptions. This could well change in the next few

decades.

IV.

In Europe, the very idea of perception emerged relatively late. Victor Caston contends

in his entry on Ancient Theories (I.1) that the early Greek thinkers did not clearly

distinguish perception from cognition, possessing only the verb ‘perceive’ to carve

out the category—in Greek as in English, this verb does not necessarily connote sense

perception. It was perhaps Plato who first attempted an analysis, saying that

perception is passive, related specifically to certain bodily organs, and outside of

rationality and thus shared with animals. This initiates a very long tradition of

distinguishing between sense-perception and discursive rationality: even today, it is

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

17

not entirely clear what part of our awareness should be attributed to the senses, and

what to learning and rational inference. It is this distinction, precisely, that is at issue

for when we speak of literally and directly seeing faces, hearing melodies, and,

(following Michotte) of apprehending causality perceptually. The developments

related above tend to shift such entities from the realm of rational inference or

learned association to that of perception.

How does perception inform us of our surroundings? Two issues dominated

the ancient debate and continue to have considerable importance today.1

The first concerns the causal influence of the object. For though it has always

been agreed that objects make themselves known by causally influencing the sense-

organs and (further downstream) the mind, the exact nature of this influence has

been hotly debated. A bright light makes me blink; a sudden loud sound startles me.

This is perception. On the other hand, the Sun makes my skin get darker; it also makes

heliotropic flowers change their orientation. It seems that these organic responses

are not perception. Why not? Like the bright light, the Sun causally evokes an organic

response. Aristotle charged his predecessors with being insufficiently mindful of the

distinction between perception and changes of the latter sort.

Aristotle himself considered perception to be the transference of form to the

sense organ without matter: for example, when we see a coloured object, the colour

is transferred to the sense organ, but without the physical substrate in which it

resides. The sense organ has, according to this view, a neutral state that is disturbed

by the reception of sensory form; when the organ is no longer in contact with the

object, it returns to its neutral state. It ceases to reflect the form of the object with

which it is causally connected, and returns to a state of receptivity to a fresh object of

perception. Perception is a state of a specific kind, which must be maintained by the

on-going causal influence of the object.

1 Brad Inwood contributed a great deal, including stretches of text, to my discussion of ancient

theories. His help was indispensible because there is no separate entry on the Epicureans and Stoics in this Handbook, and the Introduction serves partially to fill the gap.

INTRODUCTION

18

Aristotle’s form-without-matter theory appears to be modelled on the so-

called telic senses, vision and audition, which record the presence and characteristics

of distant objects and events. When we see or hear them, Aristotle wants to say

something within us takes on the colour and sound of distant objects. It is unclear

how this is meant to apply to the contact senses of touch, smell, and taste. Aristotle

posits a medium for these, and presumably he thought that form is transmitted

through the medium; in effect, this seems to subsume the contact senses to a telic

model—just as colour is transmitted through “the transparent” to the eye, so the

fuzziness of a peach is transferred to the tactile sense organ by flesh.

What could it mean to say that we receive the form of fuzziness when we feel

a peach? Is there supposed to be something in us that becomes fuzzy, though without

taking in the matter of the peach? This is implausible—even more so than to hold that

the eye takes on the colour of the peach without its matter. (As noted earlier, the

doctrine about colour is mistaken because, while it is true that the peach throws an

orange image on the retina, the only thing relevant to perception is the neural

activation caused by the image of the peach. Neural activation has no colour.)

However this may be, why is the Sun’s influence on my skin not perception in

Aristotle’s system? Perhaps, because my skin does not take on the Sun’s form in this

causal transaction, but takes on a different form, a dark colour. This in turn means

that the skin, by contrast with the sense organs, lacks the right sort of responsiveness.

Perhaps more importantly, skin does not have a neutral state or ‘perceptual mean,’ to

which it immediately returns when it is not under the influence of the Sun. We

perceive colour when the visual organ is pulled off its naturally colourless state by

receiving an object’s form of colour. When the coloured object goes out of sight, its

effect disappears; the visual organ immediately returns to the mean, and is then ready

to take on the colour of whatever else comes into view. This is not true of my skin’s

darkness when the Sun sets. Plausibly, though, my skin might perceive the Sun’s

warmth. When it is warmed by the Sun, it does take on the Sun’s form, warmth; at

night when the Sun has gone down, my skin ceases to register its warmth.

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

19

Epicureans and Stoics, causal theorists along broadly similar lines, also built

their theories of perception on the basis of their distinctive physical theories—atomic

films shed by objects entering the sense organs for the Epicureans, qualitative

changes in the organs of perception transmitted through a continuous physical

medium for the Stoics. These philosophers do not use the distinctively Aristotelian

apparatus of form and matter, but they echo Aristotle, nevertheless, by adopting a

causal model based on the telic senses.

A second issue in the ancient debate concerns the mental significance of

perception from the point of view of the perceiver. According to Plato, perception

competes with reason; according to Aristotle, it complements it. The senses “tell” us

that certain states of affairs obtain; reason and memory can equally well “tell” us that

perception is mistaken, or that it is correct. In short, perception delivers a message

that can be evaluated as true or false. Both Plato and Aristotle are thereby committed

to the view that just as reason delivers propositions for our consideration, so also

does perception—what they disagree about is the reliability and coherence of the

propositions delivered.

Supposing that perception carries propositional content, what is the nature of

the content? Both Plato and Aristotle restricted the features of which we are

perceptually aware to those that the senses are especially attuned. Colour, shape, and

pitch are properties to which the senses are peculiarly attuned; being and unity are

not. Colour, then, is a (or rather the) “proper sensible” for vision. Both Plato and

Aristotle insist, however, that the sensible qualities must come together in a central

cognitive faculty. I apprehend the pale woman singing, her hand on my shoulder: this

is a synthesis of what vision, audition, and touch tell me, a synthesis that cannot be

performed by the individual senses since, for instance, colour is not special to touch

and pressure not special to vision. It must, therefore, be performed by some facility

for sensory confluence. I recognize, moreover, that the woman’s pallor is different

from the pitch of the notes she emits. This is an act of rational differentiation.

INTRODUCTION

20

Epicurus disagreed with Plato and Aristotle about the first point; he did not

think that perception could be false. Provided that we properly focus our thoughts, he

held, perception always leads us to the truth; “falsehood and error are always located

in the opinion we add.” It is likely that he blamed the faculty of judgement for false

propositional content: if we form judgements in an appropriately receptive way, we

will not err, but if we add something to “impression,” then we risk error. (It would

have been odd for Epicurus to think that falsity, but not truth, was “located in the

opinion we add.” Most likely, this is what he thought—error is due to wrongful

insistence. It was also open to him to take the position that perception is neither true

nor false. Propositional content comes with judgement, or opinion.)

Like Plato and Aristotle, the Stoics held that, in humans, perceptual states

(which they called phantasiai, or “impressions”) convey articulable propositional

content that leads to perceptual belief when accepted or rejected. They held, in

addition, that these impressions “reveal themselves and their cause,” but here their

position is nuanced. They hold that perception is generally reliable, but acknowledge

that some impressions bring misinformation about the world. Sceptics conclude that

this impugns the entire class of impressions. The Stoics disagree. For them, an

important subset of perceptual impressions is marked by a self-validating clarity and

reliability. When I look at something attentively in good light, for example, I can be

certain that I see it as it actually is. Such “apprehensive impressions” (katalēptikai

phantasiai) are the foundation and touchstone by reference to which they believe that

we can achieve certainty about the physical world. On the other hand, a square tower

viewed from the distance looks round—this impression carries on its face its failure

to be apprehensive, because the tower looks far away and shimmery.

By taking this position, it should be noted, the Stoics align themselves with the

idea that rational belief requires certainty. They both underestimate how illusions

can take place even in the best conditions, and are also myopic with regard to the

efficacy of probable reasoning, which was recognized by the Academic sceptic,

Arcesilaus. On the other hand, they were prescient, as it turns out, in suggesting that

the perceptual given includes self-regarding reliability estimates—vision doesn’t just

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

21

tell you that a shape is concave; in many instances, it also includes an estimate of the

reliability of this attribution, depending on the goodness of the illumination, the

sharpness of resolution, and so on. RIM might suggest that estimates of reliability are

necessarily post-perceptual, but there is reason to doubt this. (Rescorla, VI.2,

discusses Bayesian models of perceptual processing, in which such estimates play a

role.)

The thinkers we have considered so far evaluate perception as true or as false.

(I suggested that Epicurus had a bit of wiggle room here.) One could deny this; one

could hold that perception, or rather sensation, is simply an effect created in

consciousness by the outside world. We draw inferences about the external world

from these effects, it could be held, but the accuracy or error of these inferences is not

to be attributed to perception itself. This attitude becomes more prominent in the

early modern period, when (as Alison Simmons notes in I.4) Descartes distinguished

between physical motions in the sense organs, the sensations occasioned by these,

and judgements that we make on the back of these sensations. Only the last of these

has propositional content. Descartes’ position on the issue of propositional content is

thus importantly different from both Epicureanism and Stoicism, though his focus on

clarity and distinctness is something he shares with both Hellenistic schools.

These questions continue into medieval philosophy as Dominik Perler

recounts in chapter I.2. He focuses on three key problems, each of which continues

and draws on ancient philosophical discussion: What is the object of perception?

What is the nature of the cognitive faculties that we need in order to apprehend these

objects? And how trustworthy are the perceptual faculties with respect to what they

reveal?

V.

Aristotle and Epicurus are in the same camp about the reliability of the senses. As we

noted above, Epicurus is the more optimistic—he does not think that there is false

perception—but Aristotle too thinks that perception leads to rational knowledge (see

especially Posterior Analytics II, 19). The Stoics too held that perception could be the

INTRODUCTION

22

foundation for knowledge, since for them, the “apprehensive impression” is, once

successfully identified, a reliable foundation for all further cognition. Plato was much

more pessimistic on this matter. According to him, the senses deliver constantly

shifting and contradictory information. They deliver a vague and confused message,

which cannot be a foundation for knowledge even where they serve as a rough and

ready guide to ordinary talk and action.

The ancient sceptics (Reed, I.3) took Plato’s pessimism to an extreme. Reacting

against Stoics, Epicureans and Aristotle alike, they sought to show that the senses are

not to be trusted. Up until now, they say—that is, up until the moment of speaking—

they have never been convinced of anything, neither by the senses nor by rational

argument. Perhaps some day they will encounter an apprehensive impression that is

so clear and distinct that it is self-validating, as the Stoics claim—but so far they

haven’t experienced anything like this. Perhaps they will, sometime in the future,

encounter a convincing argument that knowledge can be founded on the senses, but

so far all that they have experienced is doubt. The sceptical ideology prevents them

from making positive pronouncements about the reliability of the senses—they know

nothing, not even that knowledge is impossible. They parade a comprehensive

armoury of arguments against all possible claims to knowledge, but they do not affirm

the completeness of their armoury against all possible challenge. There is, of course,

a certain irony in this show of modesty.

Though scepticism never completely slipped out of view as a philosophical

tradition, it grew less important in the medieval period, with its emphasis on religious

faith. Even then, the influential Persian Muslim philosopher, al Ghazali, had

considerable affinity with scepticism, and sceptical aperçus are occasionally to be

found in such Christian thinkers as Augustine. But outright scepticism was not an

option for these thinkers in their culture. The ancient sceptics held that it was fine to

play along with religious practice as a matter of “custom,” but not as a matter of belief.

Their attitude could be expressed in this way: “Bend your head to worship: to do

otherwise, would be defy society, thereby showing epistemic arrogance. But do not

believe in God.” Such prevarication would have gone down as smoothly in medieval

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

23

centres of Christianity and Islam as it does in contemporary South Carolina or Qom.

Scepticism re-emerged forcefully much later, first in fifteenth century Italy, then

when Henri Estienne translated Sextus into Latin in 1562, and shortly thereafter in

France with the essays of Montaigne. Descartes’ methodological scepticism, which

owes much to Sextus, had a profoundly disruptive and revolutionary influence in the

early modern period. Allison Simmons says in I.4 that it led to a re-examination of

“almost all aspects of perception.”

Simmons points out, interestingly enough, that received views of perception

came under attack, at this juncture, from science. The Scientific Revolution of the 16th

and 17th centuries posited a world that was in its essence very different from the

world that perception reveals. Atoms do not, for example, have colour or smell, not

even unperceived colour and smell. If, as was increasingly popular to suppose, atoms

are the ultimate stuff of physical reality, how can the larger objects composed of

atoms have colour or smell? And what does this say about the veracity of our senses?

This disjuncture between science and perception was addressed by a

distinction between primary and secondary qualities discussed by Simmons and by

Peter Ross (in IV.2). The idea traces back to Democritus: “‘by convention sweet and

by convention bitter, by convention hot, by convention cold, by convention color; but

in reality atoms and void.’ The term ‘by convention’ is meant here to contrast with ‘by

nature.’ The idea is that things are not sweet, hot, or coloured in themselves, or by

nature. Rather, they are so relative to the perceptual response of the observer. As Ross

shows, Democritus’s idea and its successor, Locke’s distinction between primary and

secondary qualities, has many different incarnations. Ontologically, some such

distinction is needed to bridge the divide between the physical world as posited by

science, and the “manifest image” by which we initially know it.

Perhaps the most enduring legacy of scepticism is the notion of a “veil of ideas.”

Why should we refrain from accepting the evidence of our senses? One major reason

is that this evidence is held to be equivocal, much more so than it appears to the naïve

observer. Objects of different sizes or colours can create the same sense perception if

INTRODUCTION

24

they are at different distances or in different conditions of illumination; the same

thing can sound loud close up and faint further away; touch is modulated by pressure

exerted; and so on (but see Cohen, V.4, on Perceptual Constancy). According to the

sceptic, this implies that any given perceptual state betokens many different real

world situations, and thus fails to validate inference to any one of these. What is the

similarity that defines sameness of perceptual state? The standard view in both

medieval and early modern philosophy was that sameness was determined by the

ideas entertained by the perceiver. These ideas constitute an intermediary or “veil”

through which we observe the world; perception offers us certainty with regard to

the ideas, but not with regard to the world that lies beyond. The power of this doctrine

is attested to by discussions in Perler (I.2), Simmons (I.4), Paul Snowdon (I.6), Bence

Nanay (II.1), and Heather Logue (II.4). Siegel and Silins (VII.1) discuss how perception

is reason-giving with respect to belief, and the implications for scepticism.

VI.

It has now become almost commonplace to note that the philosophy of perception

suffered, until recently, from an excessive concentration on vision, which was taken

as the proxy for all of the other senses. The result is, as David Hilbert puts it, that

“vision itself, with its own peculiarities and distinctive features has a tendency to fade

from view and what we are left with is a generic sense” (III.1). As Aristotle urged, it is

necessary to bring the senses under a unified rubric. Otherwise, we will be unable to

differentiate them from other information-gathering facilities such as the immune

system. (Matthen, V.1, and Ritchie and Carruthers, III.5, take different approaches to

this problem.) However, this unified rubric is insufficiently informative about the

“peculiarities and distinctive features” of individual sense modalities. Aristotle was

aware of this. He gave a comprehensive characterization of sense perception (form

without matter, neutral state of sense organs), but he also discusses the medium,

special objects, and limitations of the individual senses separately.

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

25

In actual fact, the senses are very diverse in character.2 It helps to remember

that they are biological systems that evolved to give organisms an advantage by

providing them with the means by which to respond effectively to the challenges of

living and reproducing in surroundings that are constantly in flux. Thought of in this

way, the senses are not simply information-sinks—organs that happen to receive

ambient information at their sensory receptors, leaving their possessors to determine

how to use this information. Nor are they engineered to seek information optimal for

the organism’s pre-existing needs. Rather, they are evolved systems, with all of the

random fit to the environment that such systems display.

The evolution of sensory systems usually begins modestly with an ecologically

sensitive receptor that allows an organism to modify its behaviour to suit

circumstances. The evolution of vision, for example, begins with molecules known as

opsins, possibly derived from molecules involved in photosynthesis. These molecules

afforded primitive organisms access to information carried by light; at first, the

information available from this source is minimal—perhaps just enough to regulate

circadian rhythms. In the case of audition, evolution starts from a fluid filled chamber

that picks up vibrations from bones and other rigid structures; again, an organism

would benefit from this, miniscule though the quantity of available information would

have been. In both cases, evolution needs to add nerves that can communicate the

state of these receptors to behaviour controllers. In time, it adds facilities to refine

information collection. Given that each such step is a small advantage that a particular

population of organisms manages to gain in its local environment, it is path

dependent and unpredictable in advance how these systems evolve. The complex

utilization of information carried by light and sound that we find in more recently

evolved animals, such as mammals, is the outcome of a historically contingent

evolutionary pathway from a starting point that could have presaged different

outcomes if chance had played differently.

2 I am grateful to Jonathan Cohen, Yasmina Jraissati, and Diana Raffman for critical discussion

of this section.

INTRODUCTION

26

Vision and audition illustrate the path-dependency and contingency of these

developments. Vision receives information from light, audition from sound. The

wavelength of light closely matches the size of the molecules that make up the

everyday objects light bounces off. Consequently, light interacts with the molecules

of things that it encounters and is modified when it is reflected; reflected light is

informative about enduring characteristics of the objects. Moreover, light can be, and

is, focussed by a lens; the resulting image recapitulates the spatial distribution of

environmental sources of light (including self-luminous objects and reflecting

objects). Combining these, vision extracts from light a “map” or image of luminous

and reflecting objects together with informational content concerning certain

characteristics of these objects.

Sound is very different. First, daylight is constant and enveloping; it is a

background condition of the information that arrives at the eye. Sound, on the other

hand, comes from myriad local events, and is highly variable. (Right now, I am looking

at a calm blue sky that illuminates everything outside my window and, more diffusely,

everything inside. Sonically, however, I hear only the banging of a garbage truck,

which will shortly be replaced by silence.) Secondly, reflection scarcely (if at all)

modifies sound, so reflected sound carries information about its ultimate source, not

about the objects off which it is reflected. Finally, biology (and also, for that matter,

human engineering) has not succeeded in constructing a lens that would produce a

spatial image of the objects from which sound is reflected. (Ultrasound machines

produce such images, but they have to produce sound, or ultrasound, and capture the

returning echo. Ambient sound, which is unpredictable in direction and amplitude,

will not do. Bats, of course, do the same, as do some blind people who produce a

stream of sonic clicks, which affords them very rough echo-location.) Put all of this

together, and the result is that sound carries information about events that produce

it, but hardly any about objects off which it is reflected.

Turning now to the receptors, the frequency composition of light cannot be

exactly analysed by any biological system, while frequency composition of sound is

much more easily analysed, using mechanical resonators. The basilar membrane in

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

27

the inner ear consists of fibres, each of which has a different resonance frequency. To

exaggerate just a little, it incorporates a dedicated receptor for each and every

acoustic frequency. The analogue for light—specialized receptors for every frequency

in the visible spectrum—is not biologically feasible. The visual system computes

colour from the responses of just a few types of cells (three in the case of most

humans), each of which responds to a broad visible-frequency range. Colour vision

does little more than register total energy in these broad ranges. (More precisely, it

registers normalized differences between energy levels in these ranges. For more

about colour vision, particularly about the extraction of information from cone cells,

and the derivative character of colour perception in many situations, see Kathleen

Akins and Martin Hahn, IV.3.)

Audition extracts information about material objects by analysing the

frequency composition of the sounds they produce. It identifies voices by timbre in

speech; it identifies musical instruments and materials by the sounds they emit when

struck, plucked, or bowed. This is event based: we recognize humans by their timbre

of their voices, but only when they speak. The quality of this act of speech—somebody

reciting “Mary had a little lamb”—reveals something about the voice—that it is a rich

baritone from Lancashire—and enables us to recognize the voice when it sings

something entirely different. Vision is different; the shape and colour of the face is

directly revealed when light falls on it—these characteristics are not computed via

the character of events.

Light yields up only crude information about frequency composition—colour

is subjectively one of the more salient characteristics of vision, but frequency (colour)

information is actually surprisingly sparse by comparison to sound, and surprisingly

little used by the visual system by comparison with brightness contrasts. Primates

and birds—animals that have “good” colour vision—share many perceptual

discriminatory capacities with animals that do not possess equally good colour vision,

and typically do not use colour information in exercising these capacities. In

particular, visual perception of depth, three-dimensional shape, movement, and

INTRODUCTION

28

spatial layout are all available in black-and-white, as one can tell by looking at a black-

and-white movie.

To summarize: audition is concerned primarily with sound-producing events

(Nudds, III.2), the temporal order of these events, and properties of material objects

that can be computed from the acoustic frequency-composition of the sounds they

produce. Light, on the other hand, yields to the visual system information primarily

about the spatial configuration and distribution of objects, and their brightness

relative to other things seen at the same time. (See Jérôme Dokic, IV.4, for a rich

account of the structure of the visual representation of space.) Because vision is

dependent on constant ambient illumination, and not so much on events involving

objects, it engages more directly with the place and character of objects. The spatial

character of both vision and touch give these senses dominant roles in our

identification of particular objects. They are associated with demonstratives and

pointing—“that object,” we say, pointing with our fingers or our eyes, and this attracts

the gaze of our auditors. (How do vision and touch enable us to think about individual

objects? Imogen Dickie poses the question in VII.3.)

Vision and audition go in different directions in their engagement with the

environment because they have different information-gathering resources available

to them. Within their respective parameters, they target specialized objects,

depending on the interests of species. Both birds and humans are specialized for

identifying others of their species both visually and auditorily—humans are adept at

recognizing human faces and voices, and at understanding human speech; birds tend

to be specialized for producing and recognizing identifying song. These are “higher

level” capacities. Let’s suppose, for the sake of argument, that humans and birds are

equally good at discriminating lines and colours. It would not follow that they are

equally good at recognizing birds and humans. Birds recognize bird song, not human

speech; humans have it the other way around.

This kind of specialization holds true for other sense modalities. Consider de

Vignemont and Massin, III.3, who argue that the proper object of touch is pressure.

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

29

(They exclude temperature perception as a separate modality, as do Ritchie and

Carruthers, III.5.) Superficially, this conforms to the Aristotelian framework of proper

sensibles. Their point is that the tactile perceptions of “texture, vibration, weight,

contact, hardness, solidity” etc. “depend on the perception of pressure and tension.”

“There is no sui generis sense of texture distinct from the sense of pressure,” they

write, “for we feel the texture of a surface by feeling a spatio-temporal pattern of

pressure when stroking it.” However, de Vignemont and Massin do not think that we

go from experience of pressure to awareness of texture by inference. They are fully

aware that the sense of touch delivers such haptic properties as texture and weight

to sensory experience.

This marks a departure from the traditional articulation of the special object

view as proposed by philosophers as diverse as Aristotle and Hume—these

traditional philosophers imply that we have sensory consciousness only of the proper

sensibles such as pressure, and inferred knowledge of other properties such as

texture. What scientists began to understand with the discovery of the agnosias

(section II above) was that the ability to recognize certain complex objects of the

senses and awareness of these objects is simply provided to us by the perceptual

brain. It does not require learning or inference in the classic sense of those terms. The

specialized objects of perception are explored in Part IV of this Handbook.

Flavour perception is quite a different entity than these senses. (See Smith,

III.4, for full discussion.) Whereas vision and audition are shaped by receptors and

the kind of energy they receive, flavour is unified by its object: food and drink. The

tongue has receptors for basic tastes—sweet, sour, bitter, salty, etc. These basic tastes

differentiate many foods; the bitter warns of unhealthy toxins, and sweetness attracts

us toward energy rich foods. Flavour comprises a great deal more than these basic

tastes. Take two sherbets, one of cherry, one of raspberry. The basic tastes

differentiate them, but not by much. And why should they? Both are healthy, at least

as far as our savannah dwelling ancestors go. Both are sweet, and that is pretty much

all the information the tongue gives us. As far as flavour goes, however, the two

INTRODUCTION

30

confections are very different. They are very easy to tell apart. The puzzle is this: if the

tongue and its basic tastes do not differentiate, what does?

The additional qualities come mainly from olfactory receptors in the nasal

cavity. These receptors provide smell when we breathe and sniff. When we take

substances into the mouth, their vapours rise to the nasal cavity and flow over the

olfactory receptors. The flow is in the direction opposite to that when we breathe

through the nose. Olfaction from the mouth is known as “retronasal” olfaction. “Smells”

detected retronasally are experienced not as smells; rather, they contribute to the

flavour properties of food. They are responsible for the experience of food over and

above the basic tastes—but note that they meld in with taste and provide a unitary

experience. The “taste” of a cherry . . . no, that is just the sweetness and bit of sourness.

The flavour of cherry, that’s the whole thing. Though olfaction contributes to the

flavour of cherry, we cannot phenomenologically separate out the contribution of the

olfactory receptors. We don’t experience a smell plus a taste; we experience a unified

percept.

Flavour perception is different from vision and audition in that it does not have

information-humble beginnings. It has recruited an evolutionarily well-developed

nasal system very late in evolutionary history—humans are almost unique in their

use of retronasal olfaction—to work conjointly with another well-developed glottal

system. Olfaction is well developed in animals; so is the taste system. Flavour

perception in humans is the result of a coalition between the two. Flavour recruits

olfactory receptors to enhance a pre-existing system that already has a sophisticated

nutrition-regulation function. The result is a system that discriminates food and drink

far more finely than we need to discern what is healthy to eat and what is not. It seems

as if all possible senses get involved when we put food in our mouths—taste, olfaction,

touch (which plays a role in binding taste and olfactory qualities), the trigeminal pain

system (Smith, II.4)—and arrive together at a much finer discriminatory ability than

even the higher primates can deploy. What is the biological point? Why do we not just

make do with the system that the higher primates use? It may well be that, like

language, flavour plays a role in the highly creative socio-cultural system in which

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

31

only humans participate. Perhaps, humans are more adaptable because of their

flavour sense, and this could have contributed to their spread across the planet. Or

perhaps flavour does not simply have utility for nutrition, but has as well a social and

communicative role.

The bodily senses (Ritchie and Carruthers, III.5) provide another example of

heterogeneous sensory processes that are nonetheless highly integrated. Models

based on the external senses do not work smoothly with these. For, first, some

internal sense events seem to lack a characteristic phenomenology. Consider the

vestibular sense, which determines how we are oriented relative to gravitational

forces. Ritchie and Carruthers (III.5) suggest (though they do not come to a firm

conclusion about this) that this sense may not present us with a direct message (e.g.,

that we are upright). It may, instead, feed this information to vision, in the form of

visual field orientation, and proprioception, in the form of information about the

positioning of our bodily parts. In other words, we may sense gravitational

orientation only in terms of the orientation of objects, external and bodily. This puts

the status of the vestibular sense into question. Is it a sense modality if it does not

have a separate phenomenology?

A second source of conceptual cloudiness is that it is difficult to decide

whether the internal senses represent objective states of affairs, i.e., events or states

of the body, as opposed to presenting us merely with sensations. (Bence Nanay, II.1,

offers an overview of questions about perceptual representation.) Is an itch, for

example, a perception (or misperception) of a certain type of objective event in the

body, or is it merely a feeling? Recall the view (Simmons, I.4, and section IV above)

that perception is neither true nor false, but only a conscious event, or sensation, that

we use to make judgements about the world outside us. This view, applied to bodily

feelings, is thought by many to be unavoidable given the facts about bodily awareness.

The question is closely related to another: if pains and other bodily feelings are

perceptions of occurrences in the body, then it should be possible for this kind of

occurrence to go unperceived, just as it is possible for a sound to occur unheard. Is it

possible for me to have a pain in my finger of which I am completely unconscious?

INTRODUCTION

32

Many think that this makes no sense—pain is essentially conscious. What does this

do to the idea that pain is perception? Valerie Hardcastle discusses these questions in

IV.9. Jesse Prinz III.6 has a wide-ranging discussion of further questions about

unconscious perception. For example, it is well demonstrated that subjects fail

outwardly to respond to or inwardly to notice certain large events in their visual

fields. Should we say that they nonetheless see these events? The answer is far from

obvious.

One important feature of internal perception is that it carries a certain feeling

of ownership—the internal states of our bodies are felt as our own, and moreover our

bodies are felt as the subjects of internal states. This intimate connection between

ourselves and the things that happen in our bodies can be disrupted, but it has been

argued, notably by Wittgenstein, that it is ultimately immune to error. Mandrigin and

Thompson review the issue in IV.8, and Christopher Peacocke, II.2, takes us through

philosophically vital related issues of the first person perspective on objects of

perception.

VII.

Despite the differences among individual senses discussed in the previous section,

commonalities of function and functioning should also be noted.

It is characteristic of all of the senses that they present at least some of their

content as a continuously varying quantity. As Diana Raffman writes, VI.1: “one object

will look bluer or larger, or darker or brighter, than another, and one tone will be

louder than another, or sound more stable or more tonally centered in a given musical

key.” Plato was the first explicitly to notice this, in the Philebus. He argued that it was

an indication of the incoherence of the senses (and also of pleasure) that their content

is presented comparatively, in terms of “more and less.” It is a virtue of reason, he

argues, that it imposes absolute limits on this indefinite substrate of the more and the

less. Syllables and numbers mark well-defined absolute measures on pitch and tone

and on indefinite quantity, he claims; they are contributed by the rational mind.

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

33

Putting aside the normatively loaded claim about reason, Plato was prescient.

In speech, pitch, tone, and timbre are comparative, while phonemes are not. In the

realm of phonemes, there is “categorical perception,” by which a phoneme like /b/ or

/d/ can be recognized as the same across different speakers, despite differences

among them as to how they voice these phonemes—how loud, how high-pitched,

whether in a bass or baritone, and so on. There is a somewhat analogous situation in

colour perception too, where the named colours (blue, red, green, etc.) seem to be

sharply distinguished from one another, despite a continuously graded difference

underlying these boundaries. It is for this reason that the rainbow appears banded,

though it is actually a continuous wavelength gradient—there is a phenomenological

jump between wavelength-adjacent shades of blue and green, but no such jump

within blue or within green. Harmony might also be an instance of this: a gradient of

gradually less discordant chords abruptly jumps to harmony. Diana Raffman, VI.1,

deals with these issues; she also introduces us to the representation of similarity

relations as abstract “spaces” in which closeness represents similarity.

Another characteristic of all of the senses is that they display “constancies”

(Cohen V.4). Constancy is most apparent and has been most studied in vision. The

retinal image is a product not only of characteristics of external objects, but also of

the circumstances of viewing. A white cloth will, for example, throw a red image onto

the retina when it is bathed in red light. Yet, visual system function is to extract from

this image a message about the characteristics of the colour of the object

independently of the illumination in which it stands; the white object should look

white, and within limits, it usually does. (Akins and Hahn, IV.3, have a nuanced

discussion, relevant here, of how we come to see things as being of a colour.) The

same sort of thing is true of the other senses. In audition, a voice can be heard as

possessing constant qualities despite interfering noise from other sources; the sound

of the car is kept separate from that of your passenger’s voice. In touch, a granite

countertop is felt to be hard even when it is stroked with a soft polishing cloth, which

presents its yielding structure to the hand. Constancy is generally thought of as

revealing the orientation of the senses toward detecting properties of a stable

INTRODUCTION

34

external environment. As Cohen (V.4) writes: “it seems clear that constancy is an

absolutely fundamental aspect of perception, [which] will figure centrally in our

ultimate understanding of mind–world interaction.”

Attention is a feature of perceptual systems, the facility by means of which we

are able consciously to extract information from a perceptual state. There are many

forms of attention. In the 1980s, Anne Treisman and co-workers showed that one

form enabled “binding.” In the traditional RIM paradigm, form is extracted from

colour and brightness, and this view is intuitively plausible because it seems,

phenomenologically, that the boundaries of form and shape are in fact colour and

brightness boundaries. However, it was becoming increasingly clear by the 1980s

that form and colour were separately detected by separate parts of the brain.

Treisman’s discovery was that, having been separately detected, colour and form are

“bound” together when the subject attends to them. This shows that the fundamental

phenomenology of vision—that of the coincidence of colour and form boundaries—

is in fact a product of attention, not just of simple perception. Attention and

perception work together to produce characteristic visual appearances. John

Campbell, V.2, is on the trail of analogous synergies between perception (mainly

vision) and attention. He explores how it makes knowledge possible and how it

modifies perceptual experience.

Though, as we saw in the previous section, the senses are very different from

one another in how they process information and how they present the world around

us, there are clear communicative channels between them. Earlier we noticed that the

perception of speech and of flavour are multimodal; some say that touch is

multimodal as well. In V.3, Bayne and Spence discuss forms of multimodal perception.

(Their view is nuanced, as they argue that we might never be conscious of more than

one modality at any given time. Barry Smith also discusses multimodal perception in

III.4.) One consequence of multimodal interactions is that the modalities can

sometimes get mixed up. One such confusion is synesthesia. In a significant number

of people, perception in one modality gets expressed in another. For example, some

experience a particular colour whenever a particular musical note is played. There is

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

35

also within-modality synaesthesia: some associate individual alphanumeric

characters with individual colour, always experiencing, for instance, red when they

read ‘6’. Malika Auvray and Ophelia Deroy discuss the varieties of the phenomenon

and its proper philosophical description in V.5.

Another cross-modal “confusion,” often constructively employed, is the ability

to use one modality in place of another—for instance, to retrieve visual information

from specially arranged tactile stimulation. The phenomenon was discovered by Paul

Bach-y-Rita in the 1970s. He converted the brightness levels in a scene into a

vibratory field projected onto subjects’ backs. (Small vibrators were set to respond

proportionally to the brightness of their field positions.) The result was astonishing.

The subjects began to discern characteristically visual phenomena such as

perspective and occlusion in the scenes before them, and were able to do so using just

the tactile image projected onto their backs. This is the phenomenon called “sensory

substitution.” Can we say that Bach-y-Rita’s patients saw the scenes in front of them?

Julian Kiverstein, Mirko Farina and Andy Clark explore the ramifications of this and

other questions in V.6.

One particularly interesting feature about perception is that we are not all

equally good at it. Some wine consumers cannot tell the difference between white and

red wine; expert tasters can make fine distinctions regarding the origin and the age

of wine. (There has been a good deal of scepticism evinced on this topic lately; some

of it should be quashed by watching the documentary film “Somm,” which follows the

trials of six candidates vying for the designation of master sommelier. However that

might be, my point is quite simple—there are some who cannot make even the

simplest distinctions in this domain; others can make quite a few more.) Some

recognize in a glance the provenance of an old painting; others can barely tell whether

it is Indonesian or French. Rob Goldstone and Lisa Byrge, VII.2, argue that at least

some of these differences in discernment arise from “perceptual learning.” Perceptual

discrimination is sharpened by repeated practice and exposure. Or, as Goldstone and

Byrge write, “Perception can be learned. Experience shapes the way people see and

hear.”

INTRODUCTION

36

To end this section, a word about pictures. What is it about visual perception

that allows us to see three-dimensional scenes in two-dimensional arrays of pigment?

What is it about auditory perception that makes a musical sequence of tones bear

emotional content? Why do we find value in these seemingly ephemeral exercises of

our perceptual capacity. Dom Lopes explores these questions in VII.5.

VIII.

The two preceding sections took up questions about sensory processes. For instance,

we asked about visual processing and whether it is property based or event based;

we asked about the vestibular sense and whether it has a distinctive phenomenology.

Philosophers of perception also ask, more broadly, about the nature of perception as

a general faculty. We have touched on some aspects of this question earlier, especially

in connection with the history of the subject in sections IV and V above. Further

questions remain. What is the nature of the connection between perceiver and world

they perceive? How does perception relate to belief and the rational justification of

belief?

Let us put idealism to one side—the position that there is nothing outside

minds—and also scepticism—the position that we have no reason to believe anything

about the external world. On neither of these views, does the question arise of how

perception rationally grounds belief about the external world: for the idealist we do

not perceive an external world; for the sceptic, beliefs about the external world are

not justified by perception. The question then is this. Suppose we take it for granted

that we have, or can have, rational beliefs about the external world. How would

perception justify such a belief?

The standard view in the seventeenth century up until the beginning of the

twentieth, and among empiricists, for much longer, was that perception was directly

of a realm of ideas, or sense data, that come between us and the external world, and

only indirectly of the latter. As we saw in our earlier discussion of scepticism (section

V), indirect realism is motivated by the argument from illusion—the idea that

perception can fail. If two perceptual states are the same—say a mirage and a

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

37

veridical perception of a distant body of water—then they must have the same object.

Since the mirage fails, then, and by definition lacks an external object, the veridical

perception must lack an external object too. Both lack an external object. They must

both be directed to their common apparent object, which is internal. They are both

directed toward a sense-datum; the difference is that the veridical perception

happens to be validated by the actual presence of water on the horizon.

As Paul Snowdon relates in his entry, I.6, this seems to posit an unanalysed

psychological act-object relation and a realm of immaterial objects, sense-data. Sense-

data are decidedly queer: when I see something blue, sense-datum theorists say, my

visual state is directed at something in the mind. Things in the mind don’t have

extension or colour. They are not literally blue, or literally any other colour. This

rather mystical approach to the objects of perception did not sit well with the

increasingly dominant mid-century ethos of naturalism and materialism. A. J. Ayer

tried to wriggle out of the quandary by saying that the sense datum theory was just a

way of talking, an “alternative language,” as he put it, not a substantive ontological

proposal. Coming from a strong partisan of the theory, this seemed like a desperate

stratagem, for it is unclear how exactly the phenomenal sameness of two divergent

perceptual states can be accommodated by a simple terminological shift away from

ordinary object talk. Sense-datum theory passed out of fashion after Ayer; by 1980,

despite some revisionary attempts to revive it by two Australians, Frank Jackson and

Brian O’Shaughnessy, it was essentially gone from the scene. Some credit J. L. Austin

and Ludwig Wittgenstein for the defeat of sense-datum theory. It is equally plausible

to lay the blame at the feet of its last great proponent.

Another view of the perceptual relation originates from the works of Franz

Brentano and Edmund Husserl, treated together with the philosophy of Maurice

Merleau-Ponty by Charles Siewert in his entry on Phenomenological Approaches, I.7.

At some level, Brentano’s approach is similar to that of sense-datum theory. He

proposes that, in a way, the “referent” of perceptual states exists in the mind, and this

is just the sense-datum view. However, Brentano introduced an important analogy

between perception and linguistic affirmation. When I judge that there is a body of

INTRODUCTION

38

water on the horizon, I affirm something. There is something that I affirm, regardless

of whether it is true or false. Something of the sort holds also for perception. The

mirage of water on the horizon and the veridical perception of an actual wadi both

“affirm” the presence of water—that is what they have in common. The mirage is

false; that is how it differs from the veridical perception.

Husserl brought a sophisticated theory of meaning to Brentano’s basic insight

and was thus able to evade the idea that the sameness of perceptual states must

consist in the sameness of their “referent.” According to him, perceptual states have

a noêma in much the same way as sentences have a meaning or Fregean sense. (Nanay

II.1 shows how this approach is elaborated in contemporary analytic philosophy of

perception, which deals with perceptual representation in terms taken from analytic

philosophy of language.) This enabled him to show how two different perceptual

states could have the same noêma, but, when external circumstances change, different

external referents. According to this way of thinking, the situation of the mirage and

the wadi is analogous to the following: “Versailles is where the King of France lives,”

had the same meaning in the 18th century as it does now, but it was true then (the

circumstances made it so) but false now. This provided Husserl with a new approach

to defusing failures of perception. Perceptual states can also have different meanings

but the same external referent. This is something that both Brentano and sense-

datum theorists are unable to accommodate, but crucial to showing how perceptual

states can have external referents. Husserl rids himself of intermediate entities in

between the mind and the external world. He is a direct realist who has no need for

an intermediate realm of immaterial entities.

As Siewert tells us, there is an anti-scientistic strain in phenomenology, though

this is later much tempered by Merleau-Ponty. Brentano and Husserl both insist that

the meaning of perceptual states is directly available to perceivers. This attitude is

the underpinning of Husserl’s later notion of a Lebenswelt or “life-world”: a socially

constructed but deeply entrenched way of understanding the world given in

perception. Scientific concepts, such as those that were employed by the

psychologists of the time, were out of place in the psychological description of

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

39

perceptual states. We do not, for example, perceive ordinary things as possessing

reflectance or as emitting compression waves; rather we perceive them as coloured

and as noisy. This aspect of phenomenology has received less attention in analytic

circles than it deserves.

Two new approaches to the relationship between perception and world have

entered the field in the last few decades: “naïve realism,” or disjunctivism, which

Heather Logue writes about in II.4 and “enactive” accounts, which are dealt with in

Pierre Jacob’s entry, II.6. Disjunctivism is a return to a reference-only account of the

sort that Husserl tried to escape with his notion of noêma. A perceptual state is, on

this view, partially constituted by its object (or referent). As a consequence, a

perceptual relation that I bear to one thing is as such different from the same

perceptual relation borne to another, or to nothing. It follows that a hallucination,

which has no object, is as such different from a veridical perception, which has an

object. Perceivers may not be able to discern this difference, but it exists nonetheless.

(Hallucinations and veridical perceptions may be indiscernible, but still they share no

specific commonality; thus, they can be united only by a disjunction—hence the name,

“disjunctivism.”) Logue details four different contexts in which this position has been

advanced. John Campbell, in his entry on attention, V.2, argues that this phenomenon

can be properly understood only on a disjunctive approach.

Pierre Jacob introduces “action based accounts” in II.5—accounts which have

also been called “sensorimotor” and “enactivist.” Unlike disjunctivism, the primary

motivation for these approaches is empirical; the idea is that enactivism best makes

sense of certain experimentally determined facts about perception. This line of

thought goes back to the work of J. J. Gibson, who claimed that we perceive

“affordances,” or the possibilities for action that objects “afford” us. For example,

birds perceive branches of trees as possible places to perch; we perceive chairs and

stools in an analogous way. An even earlier antecedent is Jakob von Uexküll’s

influential notion that animals perceive the world in terms only of how it affects their

actions, which is faintly echoed in Husserl’s notion of the life-world.

INTRODUCTION

40

Enactivism is sometimes applied to specific sorts of perception, rather than

across the board. Jacob illustrates this with the perception of the actions of others.

According to one important paradigm, we observe action by covertly simulating it.

For instance, Giacomo Rizzolati and co-workers at the University of Parma have

developed a paradigm in which perceiving somebody else reaching for an object is

inwardly re-enacting oneself reaching for an object. They found, for example, that

when a monkey looked at somebody reaching for an object or grasping it, the

corresponding motor neuron, i.e., the one by which the monkey would control the

same action, is activated. By extension, we might perceive speech by inwardly

mimicking it, attribute thoughts to others by rehearsing how we would ourselves

think in their situation, and so on. Perceiving action is intimately tied up with

performing it.

IX.

How we talk about perception—how we use verbs like ‘see,’ for example—offers us

some clues about how we use it. Berit Brogaard writes about perceptual reports in

II.6. She observes that the verb ‘seem,’ which we often use non-perceptually, is an

etymological cousin of ‘see,’ and that the other perceptual verbs similarly have non-

perceptual uses. These perceptual verbs are often used “epistemically,” that is, they

are used to report beliefs, usually (but not always) beliefs that are perceptually

supported—for example, “It looks as if the lecturer is late.” (This, interestingly enough,

reports an absence—Roy Sorensen, IV.10, discusses ways of approaching “Perceiving

Nothings.”)

As we noted earlier, perception was probably not recognized as a distinct

faculty in early Greek philosophy; it was not clearly marked off in linguistic terms

then, and it still is not in modern languages. As well, we use learned associations to

describe beliefs, as Roderick Chisholm’s “comparative use” shows: “The cliff looked

like a dried-out body,” to repeat a slightly macabre example of Brogaard’s. It seems

plausible, then, to say that perception and belief are not clearly distinguished in

ordinary talk. The philosophical question of how perception relates to and rationally

OXFORD HANDBOOK OF THE PHILOSOPHY OF PERCEPTION

41

supports belief is of relatively recent origin, and should be regarded not as intuitively

founded, but the product of science and philosophical analysis. Ordinary speech does

not make a clear distinction.

If there is a rational connection between perception and belief, what does this

tell us about perception? Some say that since belief is conceptually articulated,

perception must be so as well. If my perceptual state is to rationalize the belief that

this ball is white and smooth, and if white and smooth are concepts applied to the ball,

then the perceptual state must contain these concepts too. How else would it

rationalize the belief? John McDowell affirms this connection in a particularly strong

form, holding (after Kant) that perception itself would not be possible if concepts

were not drawn upon in the “receptivity” that leads to perceptual experience. There

are doubts. Some point out that animals would not be capable of perception on this

account since they do not possess concepts. It could be argued in response that the

explicit possession of concepts may not be needed for the reception of conceptually

articulated content—there may be ways of registering the application of perceptual

concepts to the ball that do not demand this. It is well to note that McDowell’s version

of conceptual content would not be mitigated by this stipulation, since he requires

that the concepts be drawn on in receptivity itself. The nest of issues surrounding

perception and concepts are fully discussed by Wayne Wright, II.3; animal perception

is treated of by Brian Keeley in VII.4, who discusses the comparability of animal

perception to that in humans.

Let this suffice as a review of some broad issues about perception and to demonstrate

the richness of the study and its utility for philosophical inquiry. The authors of this

Handbook have produced original and searching, but at the same time introductory,

surveys of issues at the forefront. I hope that you, the reader, benefit from their efforts.

Acknowledgements I am very grateful to the Canada Research Chairs program, the

Social Sciences and Humanities Research Council of Canada, and the University of

INTRODUCTION

42

Toronto for sustained support for my research. My post-graduate fellows, Dustin

Stokes and Stephen Biggs, and my graduate students, Matthew Fulkerson and Kevin

Connolly, have been an inspiration and a source of instruction. (Kevin served as post-

doctoral fellow as well.) Peter Momtchiloff of Oxford University Press suggested this

project and offered encouragement at times when things seemed especially difficult;

Barry Smith, Ophelia Deroy, and the Centre for the Study of the Senses in London

drew me in, pushed the project forward, and introduced me to the world beyond.

Lana Kühle brought the project to completion with her invaluable editorial assistance

and taught me a lot about phenomenological approaches while she was at it.